Functional feature recognition for manipulation robot programming

A method is proposed to map a low-level CAD description of an object into a context dependent feature-based representation. This mapping is performed in two steps. The first is applied to a face-based boundary representation and produces a new model called shape feature object graph (SFOG). This model is feature oriented but independent from the application context. The second step interprets the intermediate SFOG description according to a feature library specified by the user and creates the context dependent feature-based model.<>